ChainCatcher news, according to sources, Microsoft has been working to increase both internal and third-party artificial intelligence models to support its flagship AI product Microsoft 365 Copilot, diversifying the existing OpenAI foundational technology and reducing costs.Microsoft is a major supporter of OpenAI, and this time Microsoft is seeking to reduce its reliance on the AI startup, which is different from the way Microsoft has boasted about its access to OpenAI models in recent years.Sources say that in addition to training its own smaller models (including the latest Phi-4), Microsoft is also working to customize other open models to make 365 Copilot faster and more efficient. It is reported that the aim of this move is to lower the costs of running 365 Copilot and potentially pass the savings on to end customers.

ChainCatcher news, according to IT Home, five major media outlets including the Canadian Press, CBC, Torstar, Postmedia, and the Globe and Mail have filed a lawsuit against OpenAI in the Ontario Superior Court, accusing it of extensively scraping their copyrighted content for training AI models like ChatGPT without permission and compensation. The lawsuit document spans 84 pages, demanding OpenAI to compensate for damages and to permanently prohibit unauthorized use of their content. OpenAI responded that its model training is based on publicly available data, adheres to fair use principles, and provides publishers with content attribution and opt-out options.

ChainCatcher news, a source familiar with the matter and two individuals discussing Amazon's plans, stated that Amazon has developed a new generative artificial intelligence that can handle images and videos in addition to text.The model can help customers search for specific scenes in video archives using simple text prompts, such as a game-winning basketball shot. During the development of the new model, Amazon indicated that it still hopes its internally developed AI can attract its cloud customers, thereby reducing reliance on Anthropic's AI, whose Claude chatbot is currently a popular product on Amazon Web Services.Last week, Amazon announced that it would double its investment in Anthropic to $8 billion.

ChainCatcher news, the open-source AI platform Sentient announced on X the launch of the OML 1.0 fingerprint library feature, which supports fingerprint recognition of AI models to verify the identity of AI model developers.Previous news, according to The Block, the open-source AI platform Sentient announced the completion of a $85 million seed round financing, led by Peter Thiel's Founders Fund, Pantera Capital, and Framework Ventures, with participation from Ethereal Ventures, Robot Ventures, LD Capital, Symbolic Capital, Delphi Ventures, Hack VC, Arrington Capital, HashKey Capital, Canonical Crypto, and Foresight Ventures.Additionally, according to the Web3 asset data platform RootData, Sentient is building community-contributed open-source AI models with the goal of directly competing with OpenAI. Sentient will launch activities for contributors, each with specific metrics to evaluate contributions and rewards based on these metrics.

ChainCatcher news, FLock.io announced the signing of a memorandum of understanding with Animoca Brands. Both parties plan to jointly develop AI models specifically designed for Animoca Brands, focusing on due diligence, investment, market making, and operational support. The FLock.io platform combines federated learning and blockchain technology, aiming to ensure data privacy through decentralized collaboration and promote collaborative training of AI models.FLock.io CFO Vincent stated that this collaboration will leverage FLock.io's privacy-preserving AI training infrastructure platform to develop four AI models in specific areas for Animoca Brands, while ensuring the localization of core data.

ChainCatcher news, according to TheVerge, OpenAI and Anthropic have agreed to provide the U.S. government with access to new AI models to help improve model safety. The two companies have signed a memorandum of understanding with the U.S. AI Safety Institute to allow the government to assess safety risks and mitigate potential issues. The U.S. AI Safety Institute stated that it will collaborate with its UK counterpart to provide feedback on safety improvements.It is reported that California recently passed the "Frontier AI Model Safety and Innovation Act" (SB 1047), which requires AI companies to take specific safety measures before training advanced foundational models. However, the bill has faced opposition from companies including OpenAI and Anthropic, who believe it could be detrimental to small open-source developers.

ChainCatcher news, the open-source computing platform Golem Network has announced the launch of Modelserve, a new service aimed at running AI models at scale with low costs. Modelserve is built on the Golem platform, allowing for seamless deployment and inference of AI models through scalable endpoints, ensuring efficient and cost-effective operation of AI applications. This service will drive the development of AI use cases and increase the demand for GPU providers.

ChainCatcher news, OpenAI has released its latest flagship model GPT-4o, which can perform real-time reasoning on audio, visual, and text inputs, focusing on the concept of a personalized, supernatural, ultra-low latency voice interaction assistant.According to information from the official OpenAI website and the official account on the X platform, the "o" in GPT4o stands for Omni (comprehensive), marking a step towards more natural human-computer interaction—accepting any combination of text, audio, and images as input and supporting the generation of any combination of text, audio, and image outputs. It can respond to audio input in 232 milliseconds, averaging 320 milliseconds, which is similar to human response times in conversations. In terms of English and code, it performs comparably to GPT-4 Turbo, with significant improvements in non-English language text, while the API is faster and 50% cheaper. Compared to existing models, GPT-4o excels particularly in visual and audio understanding. Text and image inputs will be launched today in the API and ChatGPT, while voice and video inputs will be rolled out in the coming weeks.

ChainCatcher message, Aptos Labs has signed a partnership with io.net to enhance access to scalable GPU resources, thereby improving artificial intelligence and machine learning capabilities.io.net and Aptos Labs will collaborate to develop a set of artificial intelligence models and inference products for future release on the Aptos network. These future products will clearly leverage the Aptos network to facilitate payments.

ChainCatcher news, the blockchain video transmission network Theta Labs announced that the Theta team is about to complete the final phase of EdgeCloud development and is ready with a range of open-source AI models that will be supported upon launch. EdgeCloud will provide instant access to critical GPU resources, including NVIDIA A100s, V100s, T4s, and more, equipped with dashboards and metrics to meet the needs of various business use cases.EdgeCloud will launch the following most popular AI models:LLMs: Mistral-8x7B, Mistral-7B, Google Gemma-7B, Meta Llama-2--7BImage Generation: Stable Diffusion XL (SDXL) Turbo, ControlNetText to Video: Stable Video DiffusionSpeech Recognition: OpenAI WhisperCode Generation: CodeLlama

ChainCatcher news, OKX Ventures announced an investment in the decentralized permissionless platform FLock for AI models and Dapps. FLock.io leverages the synergy of federated learning and blockchain technology to address the growing demand for models and potential data leakage threats, ensuring the security of model training without disclosing underlying source data, while providing fair rewards for data contribution and community collaboration."We are excited to participate in the investment in FLock. By combining technologies such as FL, ZK, and compression, FLock is building a decentralized AI platform and model community on-chain. Currently, the security issues surrounding AI data are a pain point in development. FLock allows users to provide local data for model training without leaving the data repository, thus ensuring data security and privacy. FLock's application scenarios are diverse, helping developers lower the development threshold for on-chain AI Dapps, and it is expected to become the infrastructure for AI in the Web3 world, facilitating better development in the AI sector," said Dora, founder of OKX Ventures.